Amazon Web Services has introduced native vector search for DynamoDB, letting developers store embeddings and run approximate nearest-neighbor queries directly inside the managed NoSQL database. The feature removes the need for a separate vector database and the data pipelines that kept those systems in sync.
The announcement arrived on Aug 16, 2026, and marks a shift for a database that many teams already use for application data. DynamoDB now supports vector indexes built on embeddings stored in table attributes, and developers can query them with the new SearchVectors API.
What the New Vector Search Does
Vector search is a technique for finding similar items based on embeddings, which are numerical representations of data used for semantic search. DynamoDB now stores those embeddings alongside regular application data, so teams can run filtered similarity searches and configure vector indexes without copying data elsewhere.
Developers can choose any embedding model, including Amazon Bedrock Titan Text Embeddings, Cohere Embed, or OpenAI text embedding models. To create a vector index, they specify required dimensions and a distance function. The index supports up to 4096 dimensions and offers Euclidean, Cosine, and Dot product distance functions. Inline filtering is also supported.
Vector indexes have no storage limits and scale horizontally as data grows. DynamoDB is fully serverless, so vector search scales automatically with no infrastructure to manage. The feature is available in all regions where DynamoDB currently operates, and it works with tables using either Standard or Standard-IA table class.
Removing the Old Data Pipeline
Previously, applications that wanted vector search had to copy data to a separate vector database and keep systems synchronized. That approach increased complexity and data transfer. Native vector search removes the extra data pipeline, so embeddings and application data are stored and queried in the same DynamoDB table.
The use cases are broad: agentic memory, retrieval augmented generation, recommendation engines, personalized experiences, and anomaly detection. Many databases have shipped vector support over the last couple of years, and DynamoDB is now joining that wave.
Jeff Barr, vice president and chief evangelist at AWS, made a bold claim about the feature's scale. He said it "scales as big as you want it (think trillions of vectors) while maintaining single-digit millisecond latency." That claim covers both scale and speed, though independent verification was not provided.
Cost Structure and Optimization
A vector index is billed on three dimensions in addition to standard DynamoDB charges for the underlying table. Users pay for data written into the index, data processed when searching, and data stored. All three are metered per byte and billed per GB.
Stay ahead of the AI curve
The most important updates, news, and content — delivered weekly.
No spam. Unsubscribe anytime.
Cost reduction techniques include using lower dimensions, minimal index projections, excluding embeddings from results, and selective partitioning. These approaches can significantly reduce vector search costs, according to AWS.
Some community members question whether the pricing will hold up against alternatives. A user named coinclick compared DynamoDB to S3 vector buckets, which are an alternative vector storage option. "S3 has limitless scale and will provide a consistent latency, even if the latency isn't great. I think DynamoDB will generally scale out pretty well too, with very low latency, but will probably be a lot more expensive than S3," coinclick said.
Community Reaction and Open Questions
The response has been mostly positive, though not without skepticism. Some practitioners argue AWS is "too late to the party," given how many databases already shipped vector support. Others wonder about the order of operations for attribute filters, specifically whether DynamoDB applies them before or after vector search.
Humayun Khan, a commenter, was enthusiastic. "This is a really exciting addition. Native vector search in DynamoDB could make building AI powered applications much simpler by keeping vector search and application data in one place," Khan said.
AWS has also planned an ExtendDB adapter, which is a DynamoDB-compatible adapter that will ship vector search for local development and self-managed deployments. That adapter is not yet available.
Tutorial and Next Steps
AWS published a tutorial titled "Build semantic search with native vector support in Amazon DynamoDB," written by Leonid Koren and Mo Kamioner. Koren is a principal NoSQL specialist solutions architect at AWS, and Kamioner is a senior DynamoDB solutions architect at AWS.
The tutorial shows how to build a Python semantic search app using Bedrock embeddings and DynamoDB to find research papers by meaning. Esra Kayabali, a principal solutions architect at AWS, also contributed to the feature's rollout.
The feature is available now, and teams can start experimenting with vector indexes in their existing DynamoDB tables. The question for many will be whether the convenience of keeping data in one place outweighs the cost, especially when compared to S3 vector buckets or dedicated vector databases.